← All products

Swaro icon

Your voice to text, and text back to voice — offline

Two ways, one app. Hold a shortcut and speak — your words appear right at the cursor, in any app. Or select any text and press a key to hear it read aloud in a natural voice. It all runs on your machine, so nothing is sent to the cloud.

Start free 14-day trial Buy once — $29
1 · Hold the key and speak 2 · Release — it types 3 · Select text — hear it

How it works

Both directions, one motion each

No setup ritual, no window to open — just a shortcut, in whatever app you're already using.

Speech → text

  1. 1

    Hold the shortcut & speak

    Press your global hotkey from any app and talk. A live waveform shows it's listening — transcribed on your machine, offline.

  2. 2

    Release — it types

    Let go and your words are pasted right at the cursor. Your clipboard is saved and restored.

Text → speech

  1. 3

    Select text & press a key

    Highlight anything on screen and hit the read-aloud hotkey to hear it in a natural voice — 30+ voices across 9 languages.

  2. 4

    Keep it as audio

    Prefer to listen later? Export what you typed or pasted as an MP3 from inside the app.

What it does

Dictation and read-aloud, without the cloud

  • Hold-to-talk: hold a global shortcut to record, release to transcribe and paste at the cursor.
  • Text to speech: select text anywhere, press a hotkey, and hear it read aloud in natural voices.
  • 30+ voices across 9 languages for read-aloud — English, Hindi, Japanese, Chinese, Korean, French, Italian, Portuguese and Spanish.
  • Local processing: Whisper transcription and Kokoro speech run on your machine after a one-time model download.
  • Clipboard preserved: what you had copied is saved before paste and restored after.
  • Word replacements: rules like "colon" → ":" applied automatically to transcriptions.
  • Smart keywords: say "rephrase as email" to tag a transcription for later processing.
  • Searchable history: recent transcriptions kept on disk, with a 7-day retention window.
  • Model manager: download and switch speech models from inside the app; it recommends one for your CPU and RAM.
  • Runs in the system tray with a global shortcut, out of your way until you need it.
  • Configurable start/stop chimes and a focused, dark interface.

A closer look

Inside Swaro

The real screens you'll use. Click any to enlarge.

Swaro's Record screen with the hold-to-speak prompt
Hold to speak. Press the shortcut and talk; your words land at the cursor.
Swaro's Test TTS screen with a voice picker, Speak and Export MP3 buttons
Read aloud. Pick a voice, hear any text, and export it to MP3.
Swaro's Model Manager listing speech models and their sizes
Pick your model. Small and fast for older machines, larger for top accuracy.
Swaro's dictation settings — shortcut, language and word replacements
Make it yours. Set the shortcut, language and word replacements for dictation.
Swaro's read-aloud settings — hotkey, voice, speed and pitch
Tune the voice. Set the read-aloud hotkey, voice, speed and pitch.
Swaro's About screen showing version 0.1.0 and that processing stays on your machine
On your machine. CPU-first, no GPU needed, processing stays local.

Why it's different

Built for the laptop you already have — even an old one

Most good dictation tools assume a constant cloud connection or a new, powerful computer. Swaro doesn't. It runs on a 12-year-old laptop. You choose a speech model to suit your machine: a small, fast one for older or low-power hardware, a larger one for the best accuracy on a newer machine. Every machine gets a model that works; what changes is accuracy and speed.

Private

Speech and text are processed locally. No audio is sent to a cloud transcription service to make the app work.

No GPU needed

Runs on the processor alone, so a modest laptop is enough. On Apple Silicon Macs it also uses the built-in graphics chip to go faster.

Desktop-ready

Available for Windows 10 and 11 and for macOS on Apple Silicon, with native paste on each.

Who it's for

Writers & journalists Developers & note-takers Support & sales teams People with RSI or typing strain Anyone on an older laptop Privacy-conscious professionals

The details

Under the hood

Platforms
Windows 10 and 11, and macOS on Apple Silicon
Hardware
No GPU required — runs on the processor alone, including on older laptops. On Apple Silicon Macs it also uses the built-in GPU (Metal) for extra speed. On Windows, processing currently runs on the processor.
Processing
Local, on your machine. Only the first model download and the voice setup need internet.
Speech models
14 Whisper models — from Tiny Fast (~31 MB) up to Large v3 (~3.1 GB), multilingual and English-only. The app recommends one for your hardware; smaller models run faster, larger ones are more accurate.
Voices
Kokoro text to speech (~326 MB): 30+ voices across 9 languages
Built with
Tauri 2 — a Rust core with a React/TypeScript front end
Data storage
Settings, history and models stored locally in the app's data folder; no telemetry
Licensing
Automatic 14-day trial, then an offline, device-bound license key. No license server — all checks are local.
Price
Free 14-day trial, then a one-time $29 lifetime license

Swaro is used every day by its maker. Start with the complete 14-day trial; no account or payment card is needed. See what changed in each version.

Try Swaro free for 14 days

Every feature, no card needed. If it earns its place, buy it once and keep it.

Download the free trial Buy once — $29